Particle impact is a common occurrence in numerous applications that involve handling and processing of powders. Depending on the impact details, it can have various implications for a process, e.g. it can affect the flow behaviour of powders due to kinetic energy dissipation, or influence the particle-particle and particle-substrate bonding mechanism, and consequently, the quality of the final film in coating processes such as cold spraying (CS). Thus, investigating the impact phenomenon is important for understanding and improving the efficiency of such processes. However, experimental investigation of particle impact is precarious, especially at high velocities, as the event takes place in an extremely short span of time. Therefore, numerical simulations provide a great means for the analysis of the phenomena taking place throughout impact. Discrete Element Method (DEM), Finite Element Method (FEM) and Molecular Dynamics (MD) are amongst the popular numerical methods used to date for the simulation of particle impact. However, these methods have certain limitations when dealing with the problem of impact, especially at large deformations. On the other hand, a method known as the Material Point Method (MPM) can be utilised to overcome such drawbacks. As MPM has seldom been used for the simulation of particle impact, it is adopted in the current work to carry out a comprehensive study of the impact phenomenon, especially when large deformation is concerned. Most studies on high-velocity impact processes like CS often overlook the influence of particle mechanical properties and density. Therefore, the present work considers a wide range of material properties and impact velocities to investigate their effect on the impact deformation behaviour.
To this end, MPM simulations are carried out for the impact of an elastic-perfectly plastic particle on a rigid wall. The results are analysed by focussing on variables and expressions that characterise the particle’s plastic deformation and rebound behaviour. It is observed that the plastic deformation of the particle is primarily governed by the incident kinetic energy and yield strength of the material. On the other hand, the recovery of deformation and material’s resistance to it-particularly at small deformation-are intuitively influenced not only by these factors, but also by the material’s Young’s modulus. Empirical equations are suggested for the prediction of the coefficient of restitution and the compression ratio of the particle, leveraging dimensionless groups. Subsequently, the capability of Artificial Intelligence (AI), specifically Machine Learning (ML) techniques, in identifying the underlying trends in the simulation data and refining the empirical equations is examined. Accordingly, the simulation results are introduced to a hybrid AI framework, which successfully recognises meaningful relationships, when presented with the already identified dimensionless groups. The limitations of the framework are then highlighted, and recommendations are made for further improvement. In the end, impact experiments are carried out to assess the accuracy of the numerical simulations and empirical equations. Elastic impact is first examined using elastic balls to validate the simulation predictions against experimental measurements. Elastic-plastic impact is then investigated using metal particles impacted in a custom-built impact device, with the measured compression ratio and coefficient of restitution compared to empirical predictions. Lastly, the applicability of the empirical compression ratio equation to high strain rate impacts is evaluated by depositing fine copper particles via aerosol deposition. The results confirm that the simulations accurately model the elastic impact, and the empirical equations can reasonably predict the compression ratio. However, the predicted coefficient of restitution is underestimated compared to the experimental values, though it performs better than a number of other theoretical/empirical equations. It is also found that the compression ratio at high strain rates is better predicted by a higher representative yield strength, attributed to work hardening dominating the overall deformation. The study combines numerical modelling, AI-driven analysis, and experimental validation, contributing to a deeper understanding of particle impact behaviour.
